Grok 4.7 Online | xAI Frontier Model on NetRoom
Get Started Free

Grok 4.7

Current version: 4.7 updated

xAI frontier model for coding and agents. 500K context, text and image input.

Category
Language
Modality
Text · Reasoning · Tool use
Context
500K
Released
Sep 2026
Run this model Estimate cost
Update

What's new in Grok 4.7

  • New, larger base model compared with Grok 4.6
  • 500,000-token context, text and image input
  • Reasoning effort low, medium, high and xhigh, high by default
  • Tool calling and schema-constrained structured output
  • Trained for long-running tasks and verifying its own work
Strengths

What it's the best tool for

  • 500,000-token context in a single request
  • Four reasoning effort levels, up to xhigh
  • Text and image input, text output
  • Tool calling and schema-constrained structured output
  • Built for long-running coding tasks and agent chains
  • Beats the previous Grok 4.6 release on xAI's own engineering and office benchmarks
Limitations

When to reach for something else

  • On very long prompts the bill jumps in a step: from 200 thousand tokens the higher rate applies to the whole request, not just the excess
  • The 500K context is smaller than the one million of the earlier Grok 4.3
  • Text output only: it reads images but does not generate them
  • High reasoning effort adds latency and token spend
  • Model knowledge is current only through May 2026
Sample output

How Grok 4.7 responds

Prompt
Here is a Python service file and a production crash log. Find the root cause, propose a minimal fix, re-read your own patch and tell me which edge cases it still misses. Answer as: cause, patch, risks.
Grok 4.7
Cause. The crash is not in the parser but in the cache: get_rate returns an object from a shared dict, the caller mutates it, and the next request gets an already corrupted value. The log confirms it: the first thousand requests pass, and errors start right after the first cache write. Patch. Return a copy: replace(cached) instead of cached, and make the dataclass frozen=True so a mutation fails at the point of the bug rather than ten calls later. Risks. A copy per call adds allocations on the hot path — a few percent at the current RPS, but worth measuring. Freezing breaks two places that assign fields after construction; those need to move into the constructor. The patch does not close the race between a cache refresh and a read from another thread: that needs a separate lock and is a different change.
Where teams use it

Four scenarios where it pays for itself

01
Long-running code work
Refactors, migrations and repository edits that need several steps and self-checking
02
Agent chains
Tool calling and strict response formats for pipelines
03
Knowledge work
Reports, documents and consolidation of large source sets
04
Image understanding
Diagrams, screenshots and charts as input alongside the text task
About model

More about Grok 4.7

Grok 4.7 Online — xAI's Frontier Model for Code and Agents

Grok 4.7 is xAI's September 2026 release, the top model of the Grok line for programming, agentic workflows and knowledge work. On NetRoom it runs straight in the browser: no VPN and no foreign card required.

What is inside

xAI built Grok 4.7 on a new, larger base model compared with Grok 4.6 and trained it with a longer reinforcement learning run on a harder mix of tasks, weighted toward problems that take hours to finish. By the vendor's own description the model verifies its own work more carefully and manages long context better. The context window is 500,000 tokens, input accepts text and images, and output is text. Knowledge is current through May 2026. Reasoning effort switches between low, medium, high and xhigh, with high as the default. Function calling and schema-constrained structured output are supported.

Vendor-published results

In its launch post xAI reports CursorBench 4.0 at 46.3% versus 40.4% for the previous release Grok 4.6, DeepSWE v1.1 at 71.0% on high reasoning effort, EEBench at 64.0%, Terminal-Bench 4.0 at 37.6% and HealthBench Professional at 56.7%. On GDPval the model scores 1695 Elo in xhigh mode. xAI also describes a rebuilt safeguard stack: on its own HackerBench v0.3 benchmark the model lets through 3.3% of risky dual-use prompts while keeping refusals low on legitimate security work. All figures are published by xAI, not measured by us.

Who it fits

Developers get multi-step repository edits, unfamiliar codebases, migrations and tasks where the model is expected to double-check itself. Agent builders get tool-calling chains with strict output formats. Analysts and document writers get reports and source consolidation, including diagrams and screenshots as input. For short high-volume calls such as classification or labelling, lighter models are the saner choice: a frontier model pays off on genuinely hard work.

Things to keep in mind

The 500K-token window is smaller than the one million offered by the earlier Grok 4.3. There is no image or video generation: pictures are read, not produced. And the important part about long prompts: once a prompt reaches 200 thousand tokens, the vendor bills the entire request at the higher rate rather than just the part above the threshold, so the bill jumps in a step and filling the context window just in case is a bad habit.

How to start

Sign up on NetRoom, top up your balance, pick Grok 4.7 in the model catalog and send your first prompt. Available from the browser and over the API; the current price is on the model card.

Recent changes

What changed Grok 4.7

  • + Added text model Grok 4.7 (xAI).
Full changelog →
Versions

Version history of Grok 4.7

Version Date What changed
4.7 current
  • New, larger base model compared with Grok 4.6
  • 500,000-token context, text and image input
  • Reasoning effort low, medium, high and xhigh, high by default
  • Tool calling and schema-constrained structured output
  • Trained for long-running tasks and verifying its own work

Use Grok 4.7 via the API

The same engine, straight from your code: one key and one balance for text, images, video and sound. Pay only for the requests you make.

curl
curl https://netroom.ai/api/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "x-ai/grok-4.7", "messages": [{"role": "user", "content": "Hello"}]}'

The format is OpenAI SDK compatible — point base_url at https://netroom.ai/api/v1 and use your NetRoom key.

The model id is already in the example. The full parameter reference and prices live in GET /api/v1/models and in the docs.

API documentation Get an API key

Try Grok 4.7
right now

Free access to basic models. No card, no obligations.